Does Every Treatment Arm in a Multi-Arm Trial Need the Same Sample Size?

Must Every Treatment Group Contain the Same Number of Participants?

Suppose you are designing a randomized trial comparing three educational interventions with conventional instruction. You have sufficient resources to recruit 240 participants, but you are uncertain how to distribute them across the four arms.

Should every arm receive 60 participants? Would it be better to assign more participants to the control group because the same control will be used in three comparisons? What if one intervention is considerably more expensive to deliver than the others?

Equal allocation is common because it is straightforward and often statistically efficient. However, equal sample sizes are not a universal requirement of randomized experiments.

The appropriate allocation depends on which treatment comparisons matter, how variable the outcomes are expected to be, and whether differences in cost, feasibility, or participant burden justify departing from equal allocation.

Equal Allocation Is Common, but Unequal Allocation Can Be Justified

How Allocation Ratios Affect Trial Precision and Feasibility

What Is an Allocation Ratio?

An allocation ratio describes the relative number of participants assigned to the trial arms.

In a conventional two-arm randomized trial, a 1:1 allocation assigns approximately equal numbers to treatment and control. A 2:1 allocation assigns approximately twice as many participants to one arm as to the other.

For example, a study with 180 participants and a 2:1 treatment-to-control allocation would assign approximately 120 participants to treatment and 60 to control.

In multi-arm trials, ratios can be expressed across all arms, such as 1:1:1 or 2:1:1.

The allocation ratio determines how the available information is distributed among treatment comparisons. It does not, by itself, establish whether randomization is valid.

Why Is Equal Allocation Often Preferred?

For a two-arm trial with independent observations, a continuous outcome, equal variances, and equal participant costs, equal allocation minimizes the variance of the estimated difference in group means for a fixed total sample.

To see why, consider the variance of a difference between two independent sample means.

The advantage follows from how sampling variance changes with group size. Adding participants to an already large group generally provides less improvement in the precision of the difference than adding them to the smaller group.

However, this conclusion depends on the assumptions. Different outcome variances, costs, or comparison structures may favor another allocation.

How Much Efficiency Is Lost With Unequal Allocation?

Under the simple equal-variance two-arm model, the relative efficiency of an unequal allocation compared with equal allocation can be expressed using the treatment-to-control allocation ratio.

Under this model, a 2:1 allocation requires approximately 12.5% more total participants than equal allocation to achieve the same variance of the treatment difference.

The precise sample size increase for a particular power target may differ because hypothesis tests also depend on the effect size, significance level, distributional assumptions, and degrees of freedom.

When Do Unequal Outcome Variances Favor Unequal Allocation?

Equal allocation is not generally variance-minimizing when outcome variability differs substantially between arms.

For independent groups with a fixed total sample and no difference in participant costs, the variance of the mean difference is minimized by allocating participants in proportion to the groups' standard deviations.

The difficulty is that arm-specific variances are not always known accurately before the trial.

Researchers should avoid selecting extreme allocation ratios using unstable pilot estimates. Sensitivity analyses across plausible variance assumptions can help determine whether unequal allocation offers a robust advantage.

How Do Recruitment and Treatment Costs Affect Allocation?

Participants in different arms may require different resources.

An experimental intervention might involve expensive equipment, specialist supervision, or intensive individual support. A control condition may be comparatively inexpensive.

When costs differ, the allocation minimizing variance for a fixed budget can differ from the allocation minimizing variance for a fixed number of participants.

For two independent groups with per-participant costs c_T and c_C, a simplified cost-adjusted allocation is:

This result does not mean that researchers should routinely assign fewer participants to expensive interventions. It illustrates how statistical and economic objectives can influence allocation.

The final ratio must still support the primary treatment comparison and ethical requirements.

Why Might a Multi-Arm Trial Need a Larger Control Group?

In a multi-arm trial, several interventions may share one control group.

The same control observations contribute to multiple treatment-versus-control comparisons.

Increasing the shared control sample can improve the precision of those comparisons simultaneously.

For k intervention arms with equal variances, equal treatment-arm sizes, equal participant costs, and a fixed total sample, minimizing the sum of the treatment-versus-control contrast variances yields:

This is not a universal optimal allocation. Maximizing the probability of detecting at least one effective treatment, achieving simultaneous power for all comparisons, or accommodating different intervention costs may require another ratio.

Researchers should establish the optimization objective before applying an allocation formula.

Does Unequal Allocation Affect Randomization Validity?

Unequal allocation does not inherently compromise randomization.

A properly generated randomization sequence can assign participants according to a prespecified 2:1, 3:2, or other allocation ratio.

Random assignment continues to protect against systematic allocation bias in expectation, provided the sequence is implemented appropriately and allocation concealment is maintained where relevant.

However, unequal allocation may produce different practical challenges.

Smaller arms may be more vulnerable to imprecise estimates, while extreme allocation ratios can complicate baseline balance and subgroup analyses.

Researchers should distinguish unequal allocation by design from unequal observed sample sizes caused by withdrawal or missing outcomes.

What Is the Difference Between Planned and Unplanned Unequal Sample Sizes?

Planned unequal allocation is established before participants are assigned to treatment groups.

For example, a protocol may specify a 2:1 treatment-to-control ratio because of recruitment preferences or other justified considerations.

Unplanned imbalance occurs when the observed sample sizes differ from the intended allocation.

This may arise through chance under unrestricted randomization, participant withdrawal, incomplete outcome data, or deviations from the allocation procedure.

The distinction matters because missing data may introduce bias even when the original randomization was valid.

Researchers should not treat post-randomization losses as though they were equivalent to a deliberately chosen allocation ratio.

Can Ethical Considerations Justify Unequal Allocation?

Sometimes. Researchers may prefer to allocate more participants to an intervention believed to be promising, particularly when participants have strong preferences or when the intervention provides an important prospect of benefit.

However, ethical justification requires careful consideration of clinical or substantive equipoise, uncertainty, risks, participant welfare, and the validity of the resulting evidence.

Allocating more participants to an experimental intervention does not establish that it is better than control.

Moreover, extreme unequal allocation may reduce the precision of the comparison and expose more participants to an intervention whose effectiveness remains uncertain.

Ethical and statistical objectives should therefore be evaluated together.

Does Unequal Allocation Help Recruitment?

It may. Some participants may be more willing to enroll when the probability of receiving an experimental intervention is greater.

For example, a 2:1 allocation gives participants a higher chance of receiving treatment than a 1:1 allocation.

However, whether this improves recruitment depends on the setting, intervention preferences, and participant understanding of the trial.

Researchers should not assume that unequal allocation necessarily increases enrollment.

Even if recruitment improves, the resulting statistical efficiency loss may require additional participants.

The overall benefit should be assessed using credible recruitment evidence and a design-specific sample size calculation.

How Does Unequal Allocation Affect Statistical Analysis?

Most standard methods for randomized treatment comparisons can accommodate unequal group sizes when their assumptions are satisfied.

For continuous outcomes, appropriate regression or analysis of covariance models may estimate treatment differences while adjusting for prespecified baseline variables.

For binary outcomes, generalized linear models or other suitable methods may be used.

However, unequal sample sizes can make violations of equal-variance assumptions more consequential for some conventional tests.

Researchers should use methods appropriate to the outcome distribution, variance structure, allocation, and estimand.

For multi-arm trials, the analysis should also address correlations among comparisons sharing a control and any relevant multiplicity requirements.

When Does Unequal Allocation Become Counterproductive?

Unequal allocation becomes problematic when one or more arms contain too little information to estimate their primary contrasts with adequate precision.

This may occur when an extreme allocation ratio is chosen to reduce costs or increase participants' chances of receiving an experimental intervention.

For example, assigning 90% of participants to treatment and 10% to control may leave the control estimate imprecise.

The resulting treatment contrast may be less informative than one obtained using a more balanced allocation.

How Should Researchers Choose the Allocation Ratio?

The allocation ratio should follow the primary estimand, outcome variance assumptions, recruitment and intervention costs, ethical considerations, and required statistical power.

For a simple two-arm trial, equal allocation is often a reasonable starting point.

For a multi-arm trial with several comparisons against one control, an enlarged shared control may be appropriate under certain optimization criteria.

For interventions with different variances or costs, unequal allocation may improve efficiency.

Researchers should evaluate whether the proposed allocation is feasible within the available recruitment population.

Allocation should be documented in the protocol and incorporated into the sample size calculation before recruitment begins.

Allocating Participants Across Four Educational Intervention Arms

The example demonstrates that unequal allocation can be statistically advantageous in a multi-arm setting.

It does not establish that the proposed ratio is universally optimal or that 240 participants are sufficient for the intended trial. A complete sample size calculation remains necessary.

Common Misconceptions About Unequal Allocation

How to Choose an Allocation Ratio for Your Trial

Begin by identifying the primary treatment comparisons. A two-arm superiority trial, a multi-arm shared-control trial, and a trial prioritizing direct treatment comparisons may have different optimal allocations.

Next, assess outcome variability, participant costs, recruitment feasibility, and any ethical considerations relevant to allocation.

The selected allocation should be prespecified, incorporated into the randomization procedure, and reflected in the sample size calculation.

Researchers should report the planned ratio and explain any substantial differences between planned and observed group sizes.

Before Choosing Equal or Unequal Trial Allocation

Frequently Asked Questions About Unequal Trial Arm Sizes

Allocation Should Serve the Comparison, Not Numerical Symmetry

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.